Vector engineering and delivery © The American Society of Gene & Cell Therapy original article original article

Capsid-modified AAV for Cutaneous Gene Transfer

© The American Society of Gene & Cell Therapy Tropism-modified AAV Vectors Overcome Barriers to Successful Cutaneous Therapy

Jessica Sallach1,2, Giovanni Di Pasquale3, Fernando Larcher4, Nadine Niehoff1,5, Matthias Rübsam1,5, Anke Huber1,2, Jay Chiorini3, David Almarza4,10, Sabine A Eming1,5,6, Hikmet Ulus7, Stephen Nishimura8, Ulrich T Hacker9, Michael Hallek1,2, Carien M Niessen1,5,6 and Hildegard Büning1,2

1Center for Molecular Medicine Cologne (CMMC), University of Cologne, Cologne, Germany; 2Department I for Internal Medicine, University Hospital Cologne, Cologne, Germany; 3Molecular Physiology and Therapeutics Branch, National Institute of Dental and Craniofacial Research, National Insti- tutes of Health, Bethesda, Maryland, USA; 4Centro de Investigacion Energetica Medioambiental y Tecnologica (CIEMAT) Madrid, Center for Biomedical Research on Rare Diseases (CIBERER-U714) and Department of Bioengineering. Universidad Carlos III de Madrid (UC3M), Madrid, Spain; 5Depart- ment of Dermatology, University Hospital of Cologne, Cologne, Germany; 6Cologne Excellence Center on Cellular Stress Responses in Aging-Associated Diseases (CECAD), Cologne, Germany; 7Children’s Medical Surgery, Cologne, Germany; 8Department of Pathology & Laboratory Medicine, Anatomic Pathology, UCSF, San Francisco, CA, USA; 9University Cancer Center Leipzig (UCCL), Leipzig, Germany; 10Current address: Molecular Immunology, UCL Institute of Child Health, London, UK.

Autologous human keratinocytes (HK) forming sheet and a strictly regulated spatiotemporal terminal differentiation grafts are approved as skin substitutes. Genetic engineer- program, which ultimately leads to the formation of the stratum ing of HK represents a promising technique to improve corneum, representing a water impermeable layer of avital corni- engraftment and survival of transplants. Although effica- fied cells.3 cious in keratinocyte-directed gene transfer, retro-/lenti- Insults to the barrier including skin wounding initiate a viral vectors may raise safety concerns when applied in well-coordinated process to restore skin barrier function that— regenerative medicine. We therefore optimized adeno- depending on the insult—includes cell proliferation, migration, associated viral (AAV) vectors of the serotype 2, charac- extracellular matrix deposition, angiogenesis, and skin remod- terized by an excellent safety profile, but lacking natural eling.4 This process fails in conditions of large cutaneous burns tropism for HK, through capsid engineering. Peptides, and chronic wounds. Limiting factors are the size of damage, an selected by AAV peptide display, engaged novel recep- altered, inflammatory environment and/or lack of proper angio- tors that increased cell entry efficiency by up to 2,500- genesis. Regenerative medicine has shown therapeutic promise in fold. The novel targeting vectors transduced HK with these settings. Specifically, autologous keratinocytes are grown on high efficiency and a remarkable specificity even in mixed either artificial or natural supporting scaffolds and skin sheets are cultures of HK and feeder cells. Moreover, differentiated subsequently transplanted. Engraftment as well as wound closure keratinocytes in organotypic airlifted three-dimensional can be significantly improved if the transplant secretes antiin- cultures were transduced following topical vector appli- flammatory cytokines and/or growth factors including angiogenic 5 cation. By exploiting comparative gene analysis we fur- mediators. These factors are only transiently required arguing for ther succeeded in identifying αvβ8 integrin as a target the use of nonintegrating vectors for gene delivery. receptor thus solving a major challenge of directed evo- Adeno-associated virus (AAV)-based vectors—classified as lution approaches and describing a promising candidate nonintegrating vectors—have emerged as one of the most prom- 6 3September2013 receptor for cutaneous gene therapy. ising delivery system for clinical use. They are of low immuno- genicity, can be produced at very high titers and deliver DNA Received 3 September 2013; accepted 11 January 2014; advance online vector genomes, which form episomes serving as templates for 11January2014 publication 25 February 2014. doi:10.1038/mt.2014.14 transcription.7 The prototype AAV vector is based on the serotype 10.1038/mt.2014.14 INTRODUCTION 2 (AAV2). It can be pseudotyped with capsids of other natural The skin is a highly specialized barrier that protects the organ- occurring serotypes or with engineered capsids.7 Molecular Therapy ism from environmental challenges while preventing loss of water AAV vectors have been successfully applied in over 80 clini- and small nutrients.1,2 The skin’s most outer layer, the epidermis, cal trials with an excellent safety profile.7 Clinical benefit was 22 is a stratified, multilayered epithelium consisting of the stratum reported for the treatment of a range of inherited diseases such as basale, the stratum spinosum, the stratum granulosum and the hemophilia B, Leber’s congenital amaurosis or lipoprotein lipase stratum corneum.1,2 To maintain proper protection against dehy- deficiency.6 The use of this vector system for the treatment of 5 dration and the environment, the barrier is not static but requires inherited skin diseases or in impaired wound healing, however, continuous renewal. This is achieved by a tightly controlled bal- has been hindered by the refractoriness of primary human kerati- 929 ance between lifelong self-renewal of cells in the stratum basale nocytes (HK) towards AAV-mediated transduction.8–12 939 The first two authors contributed equally to this work. Correspondence: Hildegard Büning, Center for Molecular Medicine Cologne (CMMC), University of Cologne, Robert-Koch-Str. 21, 50931 Cologne, 00may2014 Germany. E-mail: [email protected]

25February2014 Molecular Therapy vol. 22 no. 5, 929–939 may 2014 929 © The American Society of Gene & Cell Therapy Capsid-modified AAV for Cutaneous Gene Transfer

Poor transduction activity can be addressed by directing a 100 vectors towards a novel receptor that is expressed on the target cell, mediates vector internalization and induces a cellular pro- 90 gram leading to nuclear delivery of vector genomes and transgene 80 expression.7 To this end, vectors can be equipped with ligands that 70 are either noncovalently linked or genetically incorporated into 60 50 viral capsid or envelope proteins, a technology termed cell entry xpressing cells targeting or transductional targeting. 40 Also, the capsid of AAV2 tolerates genetic incorporation of 30 peptide ligands.7 It is a 60mer composed of three different viral % of GFP-e 20 proteins (VP1 (90 kDa), VP2 (72 kDa), and VP3 (60 kDa)), which 10 share most of their amino acid sequence.13 This common region 0 (“common VP3 region”) is frequently exploited for insertion of Donor 1 Donor 2 Donor 3 peptide ligands as it forms the capsid exterior.14 Of particular b interest is amino acid position 587 located at the tip of a peak at the 100 ανβ5 15 capsid’s threefold symmetry axis. Its exposed position facilitates 90 α5β1 receptor binding of inserted peptides. Furthermore, insertions at 80 HSPG this site disrupt the primary receptor-binding motif, a prerequisite 70 16–18 for approaches aiming to redirect the viral tropism. 60 e cells Rational design-based approaches to modify the viral tro- 50 pism are challenged by the limited knowledge on ligands capable 40 of directing target cell transduction. We and others have therefore

% of positiv 30 developed a high-throughput screening technique, the AAV peptide 20 display, to select in a directed evolution approach and within the physiological context of the viral capsid for appropriate ligands.19–21 10 Here, we have exploited this technology to develop AAV tar- 0 HK HeLa geting vectors with a striking tropism for primary HK. In contrast to natural occurring AAVs, these novel vectors efficiently trans- Figure 1 Lack of receptor expression renders primary human kera- duce HK in monocultures, show a remarkable specificity for HK tinocytes (HK) refractory to transduction by AAV2. (a) Primary HK in feeder layer cocultures and can be applied topically on airlifted isolated from foreskins of three different healthy donors were incubated with 5 × 103 vector genome containing particles (vg) per cell of AAV2. three-dimensional (3D) organotypic skin cultures for transduc- The AAV2 vector encoded for enhanced green fluorescent protein (GFP) tion of differentiated keratinocytes. The latter feature is not only in a self-complementary vector genome conformation. Percentage of important in basic research, but represents a central prerequisite transgene expressing cells was quantified by flow cytometry 48 hours posttransduction. Mean of three independent experiments ± SD is for future clinical applications. Finally, we established compara- shown. (b) HK and, as control, the cervix carcinoma cell line HeLa were tive gene analysis as a tool to identify receptors targeted by vectors stained with mouse antihuman αvβ5, mouse antihuman α5β1 or mouse selected in high-throughput screens. We thereby solved a major antihuman heparan sulfate proteoglycan (HSPG) antibody, followed by problem impeding the application of tropism-modified vectors a phycoerythrin-labeled IgG secondary antibody. The percentage of cells positive for HSPG or integrin expression was quantified by flow cytom- and identified a novel candidate receptor for modifying keratino- etry. Mean of three independent experiments ± SD is shown. cytes in regenerative medicine and gene therapy. HSPG-binding peptides. Aiming to select for AAV variants that RESULTS transduce HK with high efficiency and in a donor-independent Primary HK lack primary receptor for AAV2 manner, we used a different donor for each selection round and Previous studies have demonstrated that HK are poor targets for constantly increased the selection pressure by decreasing the AAV2 vectors.8–12 In line, transduction of HK, isolated from fore- number of particles from initially 1,000 vector genome containing skins of different healthy donors, did not exceed 5% (Figure 1a). particles (vg) per cell to 1 vg/cell. To unravel the cause for this refractoriness, keratinocytes were After the 5th selection round, viral genomes were isolated and analyzed for the presence of receptors required for cell trans- sequenced (Table 1). In total, nine different motifs were selected. duction by flow cytometry revealing that HK do express AAV2s’ With the exception of RSDLASL (Kera4) and PRGELAP (Kera5), all internalization receptors αvβ5 and α5β1 integrin,22,23 but lack isolates encoded peptides with a RGD tripartite motif, a hallmark of heparan sulfate proteoglycan (HSPG), which serves as attachment integrin-binding ligands. Furthermore, in three of the nine motifs, receptor (Figure 1b).24 the structure determining amino acid residue proline was located in position one and seven, and in three of the nine sequences the RGD High-throughput selection of AAV capsid variants on motif was followed by leucine-alanine or leucine-arginine. primary HK Three motifs with an overall neutral charge were picked for To redirect the tropism of AAV vectors towards HK, we per- further analyses. Specifically, we choose Kera1 (RGDTATL) formed an AAV peptide display selection (Supplementary Figure and Kera3 (RGDQQSL), because of the high frequency of their S1) with a library that had been depleted for mutants displaying occurrence, and Kera2 (PRGDLAP) because it contains the

930 www.moleculartherapy.org vol. 22 no. 5 may 2014 © The American Society of Gene & Cell Therapy Capsid-modified AAV for Cutaneous Gene Transfer

leucine-alanine motif, the potential integrin-binding domain and ratio for the AAV targeting vectors (AAV-Kera1, AAV-Kera2, and flanking proline residues, which likely stabilizes presentation of AAV-Kera3) and for AAV2 was comparable (Supplementary the inserted peptide. The three capsid variants and AAV2 were Table S1), indicating that none of the peptide insertions interfered packaged as vectors with a self-complementary vector genome with vector genome packaging.16 encoding the enhanced green fluorescent protein (GFP) controlled by the cytomegalovirus promoter. The capsid-to-genomic particle AAV targeting vectors transduce primary HK with high efficiency in a peptide-dependent manner To evaluate whether the targeting vectors indeed showed tropism Table 1 Sequence and frequency of peptide insertions selected by AAV peptide display for HK, we determined entry as well as transduction efficiency Sequence (Figure 2). For all three targeting vectors, a significantly increased Isolate of insert Frequency Net charge number of intracellular vector genomes compared with AAV2 was Kera1 RGDTATL 17 Neutral measured. The most efficient vector, AAV-Kera2, demonstrated a Kera2 PRGDLAP 2 Neutral 2,500-fold increase, followed by AAV-Kera3 (1,700-fold) and AAV- Kera3 RGDQQSL 4 Neutral Kera1 (1,600-fold). Accordingly, also the transduction efficiency Kera4 RSDLASL 3 Neutral of the novel vectors was significantly higher compared with AAV2 (Figure 2b). Incubating HK for example with 5 × 103 vg/cell resulted Kera5 PRGELAP 1 Neutral in transduction efficiencies of 54.1% ± 11.6% (AAV-Kera1), 46.7% ± Kera6 GRGDLAP 2 Neutral 16.8% (AAV-Kera2), and 47.0% ± 10.8% (AAV-Kera3), while AAV2 Kera7 RGDTASL 1 Neutral failed to transduce HK (1.2% ± 0.5%) (Supplementary Figure S2). Kera8 PRGDLRP 4 Positive Next, we performed competition assays with RGD- and RGE- Kera9 RGDQHSL 1 Positive containing peptides to prove that the RGD-containing peptide inser- Sequence is shown as one letter code. tions were responsible for the improved transduction (Figure 2c).

a b 3,000 ** AAV2 2,500 ***

units) 2,000 ** 1,500

1,000 Kera1

Intracellular vg (rel. 500

0 AAV2 Kera1 Kera2Kera3

c 90 w/o Kera2 80 RGD ** 70 RGE ** 60 * 50

xpressing cells 40 30 Kera3 20 % of GFP-e 10 0 AAV2 Kera1Kera2 Kera3

Figure 2 Adeno-associated virus (AAV) targeting vectors show a peptide-mediated superior cell entry and transduction efficiency on human primary keratinocytes. Human keratinocytes (HK) seeded on collagen-precoated plates were incubated with the three AAV targeting vectors, AAV- Kera1 (“Kera1”), AAV-Kera2 (“Kera2”), and AAV-Kera3 (“Kera3”), and AAV2, respectively. (a) For measuring cell entry efficiency, cells were incubated with 5 × 102 vg/cell. Cell transductions were stopped by extensive trypsin treatment to remove membrane-bound particles. Intracellular vector par- ticles were quantified by qPCR and normalized to the housekeeping gene Plasminogen activator (PLAT). Values obtained for AAV2 were set to 1. n = 3, **P < 0.01, ***P < 0.001. (b) HK seeded on collagen-precoated slides were incubated with indicated vectors (5 × 104 vg/cell) for 24 hours and analyzed microscopically. Original magnification: 40×. (c) HK were incubated with 5 × 103 vg/cell of AAV2 or with 7.5 × 102 vg/cell of the targeting vectors in the presence or absence of 300 µmol/l GRGDS or GRGES peptides. The percentage of transgene (GFP) expressing cells was quantified by flow cytometry 48 hours posttransduction. Mean of three independent experiments ± SD is shown. *P < 0.05, **P < 0.01.

Molecular Therapy vol. 22 no. 5 may 2014 931 © The American Society of Gene & Cell Therapy Capsid-modified AAV for Cutaneous Gene Transfer

As control, we included AAV2 with an almost sevenfold higher vg/ a 100 cell ratio than the targeting vectors. However, this was still not suf- AAV2 Kera1Kera2 Kera3 ficient to achieve a transduction efficiency above background level 80 BLM (Figure 2c). In contrast, all three targeting vectors transduced HK 2 with a vg/cell ratio of 7.5 × 10 with efficiencies of 51.0% ± 4.2% 60 (AAV-Kera3) or higher. Transduction of HK by the three targeting xpressing cells vectors was significantly reduced in the presence of RGD-containing 40 peptides, while addition of RGE-containing peptides did not com- pete with vector-mediated cell transduction (Figure 2c). The most 20 dramatic decrease in cell transduction by addition of the RGD- % of GFP-e 0 containing peptide was observed for AAV-Kera2. 10 100 1,000 2,500 5,000 (vg/cell) AAV targeting vectors demonstrate redirected b tropism 100 Having demonstrated a tropism for HK, we were interested in HepG2 studying the interaction with cell types representing potential 80 off-targets in cutaneous gene therapy approaches. To this end, we 60 compared transduction efficiencies of AAV2 and our targeting xpressing cells vectors on fibroblasts (using NIH3T3 fibroblasts), melanocytes 40 (using BLM melanoma cells) and liver cells (using HepG2 hepa- toma cells). Specifically, fibroblasts were used since this cell type 20 is not only frequently found in the dermis but because fibroblasts % of GFP-e are used as feeder cells for ex vivo keratinocyte cultures during tis- 0 10 100 1,000 2,500 5,000 sue engineering. Melanocytes were exploited as further example (vg/cell) of a specialized cell type of the epidermis and liver cells were cho- c 100 sen since AAV2 tend to accumulate in liver tissue, following both NIH3T3 local as well as intravenous application. 80 All cell lines were transduced in a dose-dependent manner by AAV2 highlighting the broad tropism of AAV vectors with wild- 60 type capsids (Figure 3). HepG2 cells showed the highest level with xpressing cells up to 77.3% ± 3.1% of transduced cells, followed by BLM cells. The 40 lowest transduction efficiency was measured for NIH3T3 cells 20

(24.3% ± 3.6%). In contrast, transduction efficiencies of AAV- % of GFP-e Kera1, AAV-Kera2, and AAV-Kera3 remained at background level 0 even at a vg/cell ratio of 5 × 103. 10 100 1,000 2,500 5,000 Since many standard HK cell cultures rely on the use of fibro- (vg/cell) blast feeder layers we aimed to evaluate whether our targeting vec- Figure 3 Adeno-associated virus (AAV) targeting vectors show target tors retain tropism for keratinocytes under the specific conditions cell selectivity. (a) The human melanoma cell line BLM, (b) the human of such cocultures. We therefore incubated cocultures of HK and hepatoma cell line HepG2, and (c) the mouse fibroblast cell line NIH3T3 NIH3T3 cells with AAV-Kera1, AAV-Kera2 and AAV-Kera3, respec- were incubated with increasing vg/cell ratios of AAV-Kera1 (“Kera1”), AAV-Kera2 (“Kera2”), and AAV-Kera3 (“Kera3”), or AAV2, respectively. tively. As indicated in Figure 4, all three targeting vectors showed a The percentage of transgene (GFP) expressing cells was quantified by remarkable preference for keratinocytes. The most striking results flow cytometry 48 hours posttransduction. Mean of three independent were obtained with AAV-Kera2, which transduced 83.8% of kera- experiments ± SD is shown. tinocytes, while only 7.1% of the fibroblasts were positive for GFP. following differentiation. While no positive cells could be detected Transduction of differentiated keratinocytes in in phosphate-buffered saline (PBS)- or AAV2-treated cultures human organotypic skin cultures (Figure 5a), strong GFP signals were seen in the most upper lay- Upon exposure to an air-liquid interface keratinocytes form a ers of AAV-Kera2– and AAV-Kera3–treated 3D skin cultures with multilayered epidermis that mimics an in vivo stratified epider- AAV-Kera2 also showing strong staining in lower keratinocyte mis including a stratum corneum.25 These 3D organotypic skin layers (Figure 5b). Only weak GFP signals were detected in the cultures serve as model system for functional testing in basic and AAV-Kera1–treated culture, suggesting that AAV-Kera1 was less applied research.25 However, at present, keratinocytes have to be efficient in transducing differentiated keratinocytes (Figure 5b). genetically manipulated prior to seeding due to inefficiency of in situ gene transfer thus limiting the use of such cultures. We AAV targeting vectors transduce primary murine therefore assessed whether our targeting vectors—when applied keratinocytes on the top layer of 16-day-old organotypic skin cultures— Besides HK, primary murine keratinocytes (MK) are also used transduce keratinocytes following multilayer formation, i.e., as targets for gene transfer but the efficiency of transfection is in

932 www.moleculartherapy.org vol. 22 no. 5 may 2014 © The American Society of Gene & Cell Therapy Capsid-modified AAV for Cutaneous Gene Transfer

Anti-Feeder GFP-expression of A GFP-expression of B a DAPI GFP Merge 26.6% 71.3% 67.3% 18.3% 1 ra AB GFP GFP Ke PBS

APC-A GFP-A GFP-A

21.2% 77.1% 83.8% 7.1% 2 ra AB GFP GFP Ke V2 AA APC-A GFP-A GFP-A

21.5% 76.7% 68.6% 11.3% 3

ra b AB GFP GFP Ke 1

APC-A GFP-A GFP-A ra Ke Figure 4 Selective transduction of human primary keratinocytes in mixed cultures. Human keratinocytes (HK) were cocultured with NIH3T3 cells and incubated with the three targeting vectors for 48 hours. Cells were harvested and stained with an anti-Feeder antibody. Whole cell population stained with anti-Feeder antibody is indicated in the left column. Percentage of transduced cells in “A” and “B” was then determined. Percentage of GFP-expressing, but anti-Feeder negative cells (population “A,” representing HK) is indicated in the middle, while percentage of GFP-expressing, anti-Feeder positive cells (population “B,” era2 representing fibroblasts) is shown in the right column. One representa- tive experiment of three is shown. general low. We therefore investigated whether our targeting vec- tors transduce primary MK. All three targeting vectors indeed transduced MK albeit with significant differences in efficiency

(Supplementary Figure S3a). The highest transduction efficiency 3K (30.9% ± 1.4%) was determined for AAV-Kera2, followed by AAV- ra Ke Kera3 (20.2% ± 2.5%) and AAV-Kera1 (15.9% ± 2.3%). Of note, in contrast to HK, which are refractory to AAV2 transduction, incubation of MK with AAV2 lead to a transduction efficiency of 12.3% ± 0.8%. Thus, while AAV-Kera1 and AAV-Kera3 showed a similar efficiency than AAV2, AAV-Kera2 was significantly more Figure 5 Adeno-associated virus (AAV) targeting vectors transduce differentiated keratinocytes in organotypic cultures. Keratinocytes in efficient in transducing MK. organotypic three-dimensional skin cultures grow airlifted on the top Next, we tested whether AAV-Kera2 is capable of transducing of a collagen matrix containing fibroblasts. 16-day-old human three- differentiated MK. We therefore applied AAV-Kera2 or PBS on the dimensional skin cultures were incubated with (a) AAV2 or (b) AAV tar- top layer of 15-day-old murine 3D organotypic skin cultures and geting vectors by dropping the vector solution into a glass ring placed on the top layer. (a) Control cultures received 1× phosphate-buffered determined transgene expression 72 hours later (Supplementary saline (PBS). Vector solution was carefully and exhaustively removed Figure S3b). While no positive cells could be detected in PBS- after 2 hours, followed by culturing of the skin equivalents for further treated cultures, GFP signals were observed in the upper kerati- 72 hours. Samples were fixed and cryo-sections were mounted in IS nocyte layer suggesting that also differentiated MK are targets for Mounting Medium DAPI. Original magnification: (a), 20×; (b), 40×. AAV-Kera2 transduction. bioinformatic approach, in identifying receptors involved in Identification of candidate receptor by comparative transduction of natural occurring viruses,26–29 we here aimed to gene analysis exploit this method to map the receptor engaged by AAV-Kera2, So far, receptors used by AAV targeting vectors selected in high- the most efficacious of our three novel targeting vectors. throughput screens remained elusive, a drawback that may We therefore incubated 51 cell lines of the NCI60 cell line hamper their clinical use due to safety issues. Prompted by the panel with AAV-Kera2 and, for comparison, with AAV2 followed success of comparative gene analysis (CGA), a microarray-based by correlation of viral transduction with the cellular expression

Molecular Therapy vol. 22 no. 5 may 2014 933 © The American Society of Gene & Cell Therapy Capsid-modified AAV for Cutaneous Gene Transfer

profile of all genes in the COMPARE database. Genes associated of αv- nor of αvβ8-blocking antibodies negatively affected cell with AAV-Kera2 transductions with a score equal or greater than transductions by AAV2. Conversely, transductions by AAV-Kera2 0.6 are shown as example in Supplementary Table S2. were remarkably inhibited by 64.3% (anti-αv, Figure 6e) or 44.3% The highest score for a cell membrane protein for AAV-Kera2 (anti-αvβ8, Figure 6f), respectively. was obtained for ITGB8 (a pearson correlation coefficient of 0.72), In support of these experimental data, sequence alignments the β8 integrin subunit. This gene positively correlated with multi- revealed a striking homology between the AAV-Kera2 peptide ple entries in the database (Table 2). In contrast, ITGB8 appeared sequence (PRGDLAP) and the binding motif (RGDLATI) of the only once in the list of genes that correlated with AAV2 transduc- high-affinity αvβ8 integrin ligand latency-associated peptide tions (pearson correlation coefficient value of 0.41). (LAP) of the transforming growth factor-β.31,32

αvβ8 integrin serves as receptor for AAV-Kera2 DISCUSSION The β8 integrin subunit solely forms heterodimers with the αv Human skin equivalents are applied in conditions of impaired integrin subunit.30 This integrin is expressed on HK (Figure 6a), healing such as chronic wounds or large skin burns. While the but not on NIH3T3, BLM or HepG2 cells (Figure 6b), the cell pro-inflammatory micromilieu represents an important barrier lines for which AAV-Kera2 did not show tropism (Figure 3). As for initial engraftment,36 inadequate perfusion hampers long-term control for this, and as a defined model cell line for subsequent persistence of the skin equivalent.37 Genetic modification of kera- experiments, we included SW480αvβ8 (Figure 6b). This cell line tinocytes represents an attractive strategy to resolve these prob- stably expresses αvβ8 integrin as well as other αv-containing lems. Furthermore, cutaneous gene therapy may also be applied integrins, while SW480, the parental cell line, expresses αv-, but to a wide range of inherited skin blistering or barrier diseases for not αvβ8 (Supplementary Figure S4a).31,32 Both cell lines were which no causal treatment is available yet.38 To push cutaneous transduced with equal efficacy by AAV2 (Supplementary Figure gene therapy towards clinical application, efficient and safe deliv- S4b). In contrast, we measured a significant higher transduction ery tools are required. Here, we developed three highly efficient efficiency of AAV-Kera2 on SW480αvβ8 compared to SW480 AAV targeting vectors (AAV-Kera1, AAV-Kera2 and AAV-Kera3) cells (Supplementary Figure S4b), which is in concordance for genetic manipulation of human primary keratinocytes (HK), with involvement of αvβ8 integrin in AAV-Kera2–mediated which transduce not only HK seeded as monolayers (Figure transductions. 2), but also showed tropism for keratinocytes in the presence αvβ8 integrin belongs to the group of integrins internalized of feeder cells (Figure 4). An important prerequisite for poten- through the clathrin pathway.33 Thus, blocking this pathway by tial clinical use of vectors across the skin barrier is the ability to chlorpromazine that inhibits assembly of clathrin lattices,34 should transduce differentiated keratinocytes. Applying our novel vectors affect AAV-Kera2–mediated transduction of HK. As control, we to airlifted differentiated organotypic skin cultures revealed that performed cell transductions in the presence of genistein, which in particular AAV-Kera2 is capable of transducing differentiated inhibits caveolin-mediated uptake.35 In the presence of genistein, keratinocytes in 3D skin cultures (Figure 5 and Supplementary the transduction rate of AAV-Kera2 was improved (Figure 6c), Figure S3b). Our study is also the first that succeeded to identify maybe as result of an enhanced cellular uptake activity induced a distinct receptor targeted by a capsid-modified vector selected by inhibition of the caveolin pathway. In contrast, and in line with in a directed evolution approach. Specifically, comparative gene our hypothesis, addition of chlorpromazine, and thus inhibition analysis (CGA) revealed αvβ8 integrin as receptor engaged by of the clathrin pathway, significantly reduced the transduction AAV-Kera2, establishing thereby CGA as strategy to map recep- efficiency of this vector (Figure 6d). tors of capsid-modified vectors, and identified αvβ8 integrin as a To provide further evidence for the involvement of αvβ8 inte- promising novel candidate receptor for cutaneous gene therapy, grin, we determined transduction efficiencies in the presence and which is of particular importance given the limited knowledge absence of αv- or αvβ8-blocking antibodies. Neither the addition on receptors that can be exploited for modification of HK. These technical improvements open up the avenue for a plethora of Table 2 The output of the COMPARE analysis new applications. More specifically, the novel vectors for example Pearson correlation represent a promising tool for application in basic research and Gene Symbol Hits coefficient preclinical models, where they are expected to promote the analy- Integrin, beta 8 ITGB8 11 0.72–0.5 sis of protein function in keratinocyte monocultures and more Glypican 4 GPC4 3 0.67–0.65 importantly in skin equivalents by improving the efficiency of gene silencing, gene addition or gene correction approaches. This Enabled homolog ENAH 3 0.63–0.53 (Drosophila) may lead to the development of novel treatment strategies includ- ing cutaneous gene therapy approaches. In this respect, AAV- Transmembrane and coiled- TMCC1 4 0.6–0.51 coil domain family 1 Kera2’s ability to transduce both primary human and murine keratinocytes (Figure 2 and Supplementary Figure S3) is an Prostaglandin-endoperoxide PTGS1 3 0.59–0.51 synthase 1 important advantage. Moreover, in a clinical setting, the novel tar- geting vectors and here again in particular AAV-Kera2 represent Kinesin family member 3A KIF3A 2 0.56–0.52 efficient delivery tools for antiinflammatory or pro-angiogenic Shown are the first six cell membrane genes that were associated with AAV- Kera2 transduction profile of the NCI60 cell panel. GeneCards symbols, number genes to HK in order to improve engraftment and survival of skin of hits, and the range of scores as pearson correlation coefficient are also listed. transplants.36,37 Since AAV-Kera2 possess the ability to transduce

934 www.moleculartherapy.org vol. 22 no. 5 may 2014 © The American Society of Gene & Cell Therapy Capsid-modified AAV for Cutaneous Gene Transfer

a

Negative control b 90 80 70 60 e cells PE-A 50 40 Anti-ανβ8 30 % of positiv 19% 20 10 0 SW480ανβ8 BLM HepG2 NIH3T3

PE-A cd* 90 90 80 80 ** 70 70 60 60 50 50 xpressing cells xpressing cells 40 40 30 30 ns 20 20 ns % of GFP-e % of GFP-e 10 10 0 0 AAV2 Kera2AAV2Kera2 ef 1.6 ns 1.6 *

1.4 1.4 units) 1.2 *** units) 1.2 *** 1 1

0.8 0.8

0.6 0.6

0.4 0.4 ansduction efficiency (rel. 0.2 ansduction efficiency (rel. 0.2 Tr Tr

0 0 AAV2 Kera2 AAV2 Kera2

Figure 6 αvβ8-integrin serves as receptor for adeno-associated virus (AAV)-Kera2. (a) Human keratinocytes (HK) seeded on collagen precoated plates as well as (b) SW480αvβ8, BLM, HepG2, and NIH3T3 cells were stained with antihuman αvβ8 antibody, followed by incubation with phycoery- thrin (PE)-conjugated secondary antibody and flow cytometric measurements. (c) HK were incubated with 5 × 103 vg/cell of AAV-Kera2 (“Kera2”) or AAV2 in the absence (black) and presence (gray) of genistein (175 µg/ml). (d) HK were incubated with 3 × 103 vg/cell of AAV-Kera2 or 5 × 103 vg/cell of AAV2 in the absence (black) and presence (gray) of chlorpromazine (16 µg/ml). (e) SW480αvβ8 cells were incubated w/o (black) or with (gray) 2 2 µg/ml of the αV-blocking antibody, followed by addition of 6 × 10 vg/cell of AAV2 and AAV-Kera2, respectively. (f) SW480αvβ8 cells were incubated w/o (black) or with (gray) 200 µg/ml αvβ8-blocking antibody, followed by addition of 7.5 × 102 vg/cell of AAV2 and 3 × 103 vg/cell of AAV-Kera2, respectively. The percentage of GFP-expressing cells (d–f) was quantified by flow cytometry 48 hours posttransduction. In a, one representative experiment of three is shown. Values of b–f represent the mean of three independent experiments ± SD. In e and f, values for nonantibody treated cells were set to 1. ns, nonsignificant, *P < 0.05, **P < 0.01, ***P < 0.001. also differentiated keratinocytes, the here-developed vectors are Our novel vectors are based on AAV2, which do not show a expected to also improve gene transfer efficiency following direct natural tropism for keratinocytes (Figure 1a). Mechanistically, injection onto the wound bed. Finally, due to the knowledge on both cell entry and postentry barriers may hamper gene transfer AAV-Kera2’s receptor, local application of the vector to αvβ8 inte- efficiency. An example for the latter was previously reported by grin-positive skin layers might possibly be used for in vivo gene Braun-Falco and colleagues who improved keratinocyte trans- transfer even in nonwounded skin. ductions by addition of MG132, a proteasome inhibitor, or of

Molecular Therapy vol. 22 no. 5 may 2014 935 © The American Society of Gene & Cell Therapy Capsid-modified AAV for Cutaneous Gene Transfer

AG1478, an epidermal growth factor receptor protein tyrosine amongst them HK and MK.9 The vectors, like ours, encoded for kinase (EGFR-PTK) inhibitor.8 Both drugs impair degradation GFP controlled by the cytomegalovirus promoter in a self-com- of incoming AAV2 particles by directly blocking proteasomal plementary vector genome conformation. In contrast to our study, degradation (MG132) or by impairing EGFR-PTK–mediated however, they performed all experiments with a vg/cell ratio of phosphorylation of tyrosine residues of the AAV capsid, which 105. Of the nine serotypes, AAV6 was the most efficient vector promotes ubiquitination and thus degradation.39,40 transducing 80% of HK, followed by AAV1 (74%) and AAV3 Our study is indicative of the presence of an additional barrier, (15%). Importantly, using a 20-fold lower vg/cell ratio, i.e., 5 × 103 namely, the lack of HSPG (Figure 1b). This cell surface molecule is vg/cell, we reached transduction efficiencies comparable to the a component of the extracellular matrix and serves as attachment most efficient serotype (AAV6), namely up to 83.8% (Figure 4). receptor for AAV2.24 Site-directed mutagenesis of the HSPG bind- The same was true for MK, for which AAV1 is reported to be ing motif16,17 or presence of heparin, a soluble analogue of HSPG, the most efficient serotype.9 Again, using a 20-fold lower vg/cell during vector application renders AAV2 noninfectious24 revealing ratio an equal transduction efficiency was reached (AAV1: 29% the importance of HSPG for cell transduction. In order to explain with a vg/cell ratio of 10e59 vs. AAV-Kera2: 30.9% ± 1.4% with a this dependency, Asokan and colleagues postulated a conforma- vg/cell ratio of 5 × 103 (Supplementary Figure S3a)). Thus, with tional change of the capsid—induced by its binding to HSPG— the here-developed vectors significantly lower particle numbers that is required for binding to the internalization receptors.22 So have to be applied to obtain transduction efficiencies similar to far, two different receptors for cell entry have been reported, α5β1 AAV1 and AAV6, the so far most efficient serotypes for primary and αvβ5 integrin,22,23 which—in contrast to HSPG—are both keratinocytes. present on HK (Figure 1b). Since each target cell type expresses a plethora of cell surface The AAV peptide display technology represents a powerful molecules, potentially serving as receptor, it is technically impos- tool to screen for viral variants that overcome both, pre- and sible to use high-throughput screens like the AAV peptide display postentry barriers, since this evolutionary approach selects for for the selection of mutants specifically targeting a certain, before- mutants capable of fulfilling the whole viral lifecycle,i.e. , gener- hand chosen receptor. Due to the limited size of the selected pep- ate progeny following target cell infection.19–21,41–44 For selection tide, it is also extremely unlikely to decipher the targeted receptor of AAV variants targeting HK, we used our previous-described by a common data bank search. Thus, characterization of a spe- library of mutants displaying 7-mer random peptides at amino cific receptor involved in retargeting represents an important so acid position 587.20 Using this position for peptide insertion far unresolved challenge. To this end, we here exploited CGA as a results in natural receptor-blinded mutants as the two main resi- novel strategy to identify the receptor targeted by the PRGDLAP dues of the HSPG binding motif, R585 and R588, become sepa- peptide of AAV-Kera2, the targeting vector with the most striking rated.16–18 Compared to previous AAV peptide display screens, tropism for HK. This screen revealed ITGB8, the integrin subunit we here implemented three technical improvements, which we β8, as candidate. β8 is unique as it is solely described as heterodi- believe are critical for selection of efficient targeting vectors mer with αv.30 The only high-affinity ligand of αvβ8 integrin is the with a redirected tropism: (i) the use of different donors in each LAP of transforming growth factor-β,32 which becomes thereby selection round to avoid bias for receptors expressed in a donor- sequestered, a process crucial for maintaining epithelial homeo- specific manner, (ii) continuous increase of the selection pressure stasis.32,46 The binding of LAP to αvβ8 integrin is mediated by a by lowering the vg/cell ratio aiming to select for the fittest vari- RGDLATI motif located in the C’-terminal part of LAP, specifi- ants and (iii) the use of a library that had been depleted for AAV cally at amino acid positions 244–250. Our selected peptide is a variants displaying HSPG binding peptides. Although HK lack truncated version of this ligand as it contains the RGDLA motif HSPG expression, the latter step was included since mutants dis- flanked by two proline residues, which may stabilize this motif playing such peptides like AAV2 tend to show a broad tropism45 when displayed at the tip of the capsid peaks. Experimental evi- and to accumulate in liver and spleen,18 which should be avoided dence that αvβ8 serves as receptor for AAV-Kera2 is provided by in cutaneous gene therapy. the presence of this integrin on the target cells (Figure 6), the tro- Sequencing of the viral progeny revealed selection of nine pism of AAV-Kera2 for cells expressing αvβ8 integrin (compare different peptide motifs (Table 1), three of which were chosen Figure 6a,b with Figures 2 and 3; Supplementary Figure S4) and for further analyses based on the frequency of their occurrence the significant inhibition of cell transduction in the presence of (Kera1 and Kera3) or on the unique combination of prominent αv- or αvβ8-blocking antibodies (Figure 6e,f). features (Kera2). All three targeting vectors (AAV-Kera1, -Kera2, In summary, we here report on the development of three and -Kera3) not only showed a donor-independent impressive AAV2-based vectors for keratinocyte-directed gene transfer increase in entry and transduction efficiency compared with approaches that overcome the refractoriness of keratinocytes by AAV2 (Figure 2 and Supplementary Figure S2), but also—in targeting novel receptors. All three vectors showed tropism for line with our hypothesis—a remarkable redirection of the tropism keratinocytes in mono- and mixed cultures and for differentiated towards HK (Figures 2–4) arguing for the efficacy of our selection keratinocytes in organotypic 3D cultures, while transduction of strategy. off-target cell types remained at background level. Furthermore, Comparison with serotypes others than AAV2 further we report on the first successful approach to identify candidate highlights the potency of the here-developed vectors. Ellis and receptors engaged by tropism-modified AAV vectors, which as colleagues compared nine different natural AAV serotypes consequence revealed αvβ8 integrin as a promising candidate (AAV1-AAV9) on 34 different cell lines and primary cell types receptor for cutaneous gene therapy.

936 www.moleculartherapy.org vol. 22 no. 5 may 2014 © The American Society of Gene & Cell Therapy Capsid-modified AAV for Cutaneous Gene Transfer

MATERIALS AND METHODS Quantification of entry and transduction efficiency. For determining Cell lines and primary cells. The human embryonic kidney cell line entry efficiency HK, seeded at 70–80% confluency in collagen precoated HEK293 (ATCC number: CRL-1573), the human cervix carcinoma cell wells, were incubated with the indicated vg/cell ratio for 30 minutes on line HeLa (ATCC number: CCL-2), the human hepatoma cell line HepG2 ice and then shifted to 37 °C and 5% CO2. Cells were harvested 1 hour (ATCC number: HB-8065), the human adenocarcinoma cell line SW480 posttransduction (p.t.). The cells were washed and membrane-bound vec- and its derivative SW480αvβ8 (ref. 31) as well as the murine fibroblast cell tor particles were removed by trypsin treatment.45 Cell pellet was washed line NIH3T3 (ATCC number: CRL-1658) were maintained in Dulbecco’s twice with 1× PBS and total viral DNA was isolated (DNeasy Tissue Kit, modified Eagle’s medium (DMEM) with GlutaMAX-I (Invitrogen, Qiagen). Relative quantification of vector genomes (GFP) and reference Karlsruhe, Germany). The human melanoma cell line BLM (kindly pro- gene (PLAT) was performed by qPCR (Roche Diagnostics).45 The speci- vided by Cornelia Mauch, University of Cologne, Germany) and the NCI60 ficity of target and reference gene amplification was confirmed by melt- cell line panel were maintained in RPMI-1640 medium with GlutaMAX-I ing-curve analysis. GFP values were normalized to PLAT levels using the (Invitrogen). All media were supplemented with 10% fetal calf serum LightCycler 480 software 1.5 (Roche Diagnostics). (Invitrogen), 100 IU/ml of penicillin (Invitrogen) and 100 µg/ml of strep- For determining transduction efficiency, HK or indicated cell lines tomycin (Invitrogen). The medium of SW480αvβ8 and of the NCI60 cell were incubated with the indicated vg/cell ratio for 48 hours, while line panel additionally contained 4 µg/ml of puromycin (Invitrogen) and MK were incubated with 5 × 103 vg/cell for 72 hours, followed by flow L-Glutamine 2 mmol/l (Invitrogen), respectively. cytometry measurements (BD FACS Calibur). A minimum of 10,000 cells HK were isolated from human foreskin of healthy donors after was counted for each sample. Background fluorescence was set to 1%. For informed consent (ethics committee vote: 12–163). Foreskin was mixed cultures, HK and NIH3T3 fibroblasts were seeded at a ratio of 1:1. separated manually into dermis and connective tissue. Dermis was Briefly, NIH3T3 cells were seeded in a collagen precoated 24-well plate minced into small pieces, transferred into Dispase II solution (10 mg/ml and cultivated for 5 hours. Medium was carefully aspirated and HK were in DMEM/GlutaMAXTM-I (Invitrogen)) and incubated for 24 hours at 4 added. The number of HK were determined 24 hours later and mixed °C. Epidermis was detached from the dermis, followed by cell separation cultures were incubated with 5 × 103 vg/HK for 48 hours at 37 °C. Cells through trypsin/EDTA treatment. The suspension was filtered, pelleted were harvested, stained with anti-Feeder antibody (monoclonal, APC by low speed centrifugation and cell pellet was resuspended in CnT Basal conjugated mouse antihuman, Miltenyi, Bergisch Gladbach, Germany) Medium 1 (CELLnTEC Advanced Cell Systems AG, Bern, Switzerland). and GFP-positive as well as APC-positive cells were determined by flow Primary keratinocytes were isolated and cultured in minimal Ca2+ cytometry as described above. medium (50 µmol/l Ca2+) as described.47 Competition and inhibition experiments. HK, seeded in collagen pre- Antibodies and evaluation of cell surface expression of integrins and coated wells, were incubated with chlorpromazine at a final concentration HSPG. For determining cell surface expression of integrins and HSPG, of 16 µg/ml or with genistein at a final concentration of 648 µmol/l for 30 respectively, the following antibodies were used: mouse antihuman αvβ5 minutes at 37 °C or left untreated. Vectors were added at the indicated vg/ antibody (MAB1961, Millipore, Schwalbach, Germany), mouse antihu- cell ratio to drug-treated or -untreated cells. Drug treatment was stopped man α5β1 antibody (MAB1999, Millipore), mouse antihuman αv anti- at indicated time points p.t. by washing, followed by extensive treatment body (MAB1953, Millipore), mouse antihuman αvβ8 antibody (37E5,)31 with trypsin and reseeding of cells in fresh medium again on precoated and mouse antihuman heparan sulfate delta (USBiological, Salem, MA). plates. For peptide competition experiments, HK were incubated with Cells were harvested by mechanically scraping, pelleted and resuspended 300 µmol/l GRGDS or GRGES peptides or left untreated for 15 minutes in 1× PBS. Indicated primary antibody was added and cells were incubated at 37 °C. Subsequently, vectors were added at the indicated vg/cell ratio. 15 minutes on ice followed by a washing step. Secondary antibody (goat At 4 hours p.t. cells were harvested and reseeded as described before. antimouse IgG, Southern Biotech, Eching, Germany) was added to cell For the antibody blocking experiments cells were either incubated with suspension in 1× PBS. Following incubation on ice for 15 minutes and a 2 µg/ml αv-blocking antibody (Millipore) for 15 minutes at 37 °C or further washing step, cells were analyzed by flow cytometry. A minimum with 200 µg/ml of αvβ8-blocking antibody31 for 30 minutes on ice, fol- of 10,000 cells was counted for each sample. Unstained cells and cells that lowed by vector incubation. Cells incubated with vectors in the absence of were only treated with secondary antibody were used as controls. antibodies served as control. Vectors were removed after 4 hours (αv) or 30 minutes­ (αvβ8). Transduction efficacy was determined 48 hours p.t. by Selection of AAV capsid mutants on primary HK. AAV peptide display flow cytometry. selection on HK was performed using a library of AAV capsid insertion mutants depleted for mutants capable of binding to HSPG by heparin affin- Correlation of AAV-Kera2 transduction with NCI60 cell panel microarray ity chromatography.18 Selection pressure was increased from round 1 to 5 gene expression pattern. Fifty-one cell lines of the NCI60 panel were trans- by reducing the vg/cell number from 1,000 to 1. Following the 5th selec- duced with serially diluted AAV2 or AAV-Kera2 vectors. 48 hours p.t. cells tion round, viral progeny was isolated and sequenced (Qiagen, Hilden, were harvested and analyzed by flow cytometry to quantify GFP-positive Germany).20 The AAV helper plasmids pRC-Kera1, pRC-Kera2 and pRC- cells using a BD FACSArray. The relative transduction efficiencies of the 51 Kera3 were obtained by subcloning the BsiwI/SnaBI fragment of the AAV cell lines were expressed as a pattern which we then used as a seed file for genome containing the respective insert sequences (Table 1) into pRC-K17 comparison to the patterns in the Developmental Therapeutics Program from which the corresponding region had been removed. (DTP), National Cancer Institute (NCI) database, using the COMPARE algorithm (http://dtp.nci.nih.gov/compare-web-public_compare/login. Production of recombinant AAV vectors. HEK293 cells were seeded at 80% do). This algorithm determines the similarity of patterns between the given 48 confluency and cotransfected with a total of 37.5 µg of pRC, pscAAV/ “seed” and others within a database by creating a scalar index of similarity 49 50 EGFP, and pXX6 for AAV2 production. For production of targeting expressed quantitatively as a pearson correlation coefficient. The output vectors, pRC was exchanged for pRC-Kera1, pRC-Kera2, and pRC-Kera3, of the COMPARE program is a rank-ordered list (by pearson correlation respectively. Forty-eight hours posttransfection, cells were harvested, lysed coefficient) of the most highly correlated patterns from the database. and purified by iodixanol density step gradient centrifugation.49 Genomic particle titers were determined by qPCR (LightCycler System, Roche Transduction of organotypic human 3D skin cultures. A dermal equiva- Diagnostics, Mannheim, Germany) using transgene specific primers.45 The lent consisting of a collagen gel populated with NIH3T3 fibroblasts was capsid titer was determined by enzyme-linked immunosorbent assay using cultivated in DMEM/GlutaMAXTM-I medium supplemented with 10% the AAV2 capsid-specific antibody A20 (Progen, Heidelberg, Germany). fetal calf serum (Invitrogen), 100 U/ml penicillin (Invitrogen), 100 µg/

Molecular Therapy vol. 22 no. 5 may 2014 937 © The American Society of Gene & Cell Therapy Capsid-modified AAV for Cutaneous Gene Transfer

ml streptomycin (Invitrogen), 10 µg/ml Transforming Growth Factor-α 9. Ellis, BL, Hirsch, ML, Barker, JC, Connelly, JP, Steininger, RJ 3rd and Porteus, MH (2013). A survey of ex vivo/in vitro transduction efficiency of mammalian primary cells (Sigma-Aldrich, Taufkirchen, Germany), 10 µg/ml EGF (Sigma-Aldrich) and cell lines with Nine natural adeno-associated virus (AAV1-9) and one engineered and 50 µg/ml sodium L-ascorbate (Sigma-Aldrich). In case of the organo- adeno-associated virus serotype. Virol J 10: 74. typic murine 3D skin culture, the collagen gel populated with NIH3T3 10. Gagnoux-Palacios, L, Hervouet, C, Spirito, F, Roques, S, Mezzina, M, Danos, O et al. (2005). Assessment of optimal transduction of primary human skin keratinocytes by fibroblasts was cultivated in DMEM/Ham’s F12 medium (Sigma-Aldrich) viral vectors. J Gene Med 7: 1178–1186. supplemented with 10% fetal calf serum Gold (PAA, Coelbe, Germany), 11. Keswani, SG, Balaji, S, Le, L, Leung, A, Lim, FY, Habli, M et al. (2012). Pseudotyped adeno-associated viral vector tropism and transduction efficiencies in murine wound 100 U/ml penicillin (Invitrogen), 100 µg/ml streptomycin (Invitrogen), 2 healing. Wound Repair Regen 20: 592–600. mmol/l L-Glutamine (Life Technologies GmbH, Darmstadt, Germany), 12. Petek, LM, Fleckman, P and Miller, DG (2010). Efficient KRT14 targeting and 10 µg/ml EGF (Sigma-Aldrich), 50 µg/ml sodium L-ascorbate (Sigma- functional characterization of transplanted human keratinocytes for the treatment of epidermolysis bullosa simplex. Mol Ther 18: 1624–1632. Aldrich), 5 µg/ml Insulin (Sigma-Aldrich), 0.5 µg/ml Hydrocortisone 13. Kronenberg, S, Kleinschmidt, JA and Böttcher, B (2001). Electron cryo-microscopy and (Sigma-Aldrich), 1 × 10-10 mol/l cholera toxin (Sigma-Aldrich), and 0.2 image reconstruction of adeno-associated virus type 2 empty capsids. EMBO Rep 2: 997–1002. mmol/l CaCl2. Keratinocytes were seeded in glass rings on top of the col- 14. Büning, H, Bolyard, CM, Hallek, M and Bartlett, JS (2011). Modification and labeling lagen gel submerged in medium for 24 hours at 37 °C. Glass rings were of AAV vector particles. Methods Mol Biol 807: 273–300. removed, medium was changed every other day with the liquid level just 15. Xie, Q, Bu, W, Bhatia, S, Hare, J, Somasundaram, T, Azzi, A et al. (2002). The atomic structure of adeno-associated virus (AAV-2), a vector for human gene therapy. Proc sufficient to contact the collagen gel, leaving the keratinocyte layer exposed Natl Acad Sci USA 99: 10405–10410. to air. To transduce HK or MK, a sterile glass ring was put on top of the 16. Kern, A, Schmidt, K, Leder, C, Müller, OJ, Wobus, CE, Bettinger, K et al. (2003). Identification of a heparin-binding motif on adeno-associated virus type 2 capsids. keratinocyte layer and filled with medium containing the indicated vec- J Virol 77: 11072–11081. tor solution or 1× PBS. Two hours p.t., vector containing solution or 1× 17. Opie, SR, Warrington, KH Jr, Agbandje-McKenna, M, Zolotukhin, S and Muzyczka, N (2003). Identification of amino acid residues in the capsid proteins PBS was carefully removed. Organotypic cultures were cultivated for fur- of adeno-associated virus type 2 that contribute to heparan sulfate proteoglycan ther 72 hours at 37 °C and subsequently fixed in 2% paraformaldehyde in binding. J Virol 77: 6995–7006. 3.5% sucrose/1× PBS for 30 minutes at room temperature and embedded 18. Perabo, L, Goldnau, D, White, K, Endell, J, Boucas, J, Humme, S et al. (2006). Heparan sulfate proteoglycan binding properties of adeno-associated virus retargeting mutants in Tissue Tek (Sakura Finetek, Zoeterwoude, Netherland). Cryo-sections and consequences for their in vivo tropism. J Virol 80: 7265–7269. (5 µm) were mounted by IS Mounting Medium DAPI (Dianova, Hamburg, 19. Müller, OJ, Kaul, F, Weitzman, MD, Pasqualini, R, Arap, W, Kleinschmidt, JA et al. (2003). Random peptide libraries displayed on adeno-associated virus to select for Germany) and analyzed microscopically. targeted gene therapy vectors. Nat Biotechnol 21: 1040–1046. 20. Perabo, L, Büning, H, Kofler, DM, Ried, MU, Girod, A, Wendtner, CM et al. (2003). SUPPLEMENTARY MATERIAL In vitro selection of viral vectors with modified tropism: the adeno-associated virus Figure S1. Schematic overview of AAV peptide display selection. display. Mol Ther 8: 151–157. 21. Varadi, K, Michelfelder, S, Korff, T, Hecker, M, Trepel, M, Katus, HA et al. (2012). Figure S2. Transduction efficiency of targeting vectors on primary Novel random peptide libraries displayed on AAV serotype 9 for selection of human keratinocytes. endothelial cell-directed gene transfer vectors. Gene Ther 19: 800–809. Figure S3. Transduction efficiency of targeting vectors on primary 22. Asokan, A, Hamra, JB, Govindasamy, L, Agbandje-McKenna, M and Samulski, RJ (2006). Adeno-associated virus type 2 contains an integrin alpha5beta1 binding murine keratinocytes. domain essential for viral cell entry. J Virol 80: 8961–8969. Figure S4. Presence of αvβ8 integrins improves cell transduction. 23. Summerford, C, Bartlett, JS and Samulski, RJ (1999). AlphaVbeta5 integrin: a co- Table S1. Packaging efficiency of AAV targeting vectors. receptor for adeno-associated virus type 2 infection. Nat Med 5: 78–82. 24. Summerford, C and Samulski, RJ (1998). Membrane-associated heparan sulfate Table S2. Genes associated with AAV-Kera2 transduction. proteoglycan is a receptor for adeno-associated virus type 2 virions. J Virol 72: 1438–1445. ACKNOWLEDGMENTS 25. Asselineau, D, Bernhard, B, Bailly, C and Darmon, M (1985). Epidermal morphogenesis and induction of the 67 kD keratin polypeptide by culture of human This work was supported by grants from the Research Priority Program keratinocytes at the liquid-air interface. Exp Cell Res 159: 536–539. 1230 (SPP1230) “Mechanisms of gene vector entry and persistence” 26. Di Pasquale, G, Davidson, BL, Stein, CS, Martins, I, Scudiero, D, Monks, A et al. (2003). of the Deutsche Forschungsgemeinschaft (DFG) to H.B. (BU1310/1– Identification of PDGFR as a receptor for AAV-5 transduction. Nat Med 9: 1306–1312. 27. Kondratowicz, AS, Lennemann, NJ, Sinn, PL, Davey, RA, Hunt, CL, Moller-Tank, S et 2) and from the Center for Molecular Medicine Cologne (CMMC) al. (2011). T-cell immunoglobulin and mucin domain 1 (TIM-1) is a receptor for Zaire to H.B. (B1). F.L. was supported by grants from Instituto de Salud Ebolavirus and Lake Victoria Marburgvirus. Proc Natl Acad Sci USA 108: 8426–8431. Carlos III (PI11/01225) and Comunidad de Madrid (S2010/BMD- 28. Quinn, K, Brindley, MA, Weller, ML, Kaludov, N, Kondratowicz, A, Hunt, CL et al. (2009). Rho GTPases modulate entry of Ebola virus and vesicular stomatitis virus 2359), S.A.E. by the CMMC (C1.1), S.N. by the National Institutes pseudotyped vectors. J Virol 83: 10176–10186. of Health (HL113032, NS044155) and M.H. by the European Union 29. Weller, ML, Amornphimoltham, P, Schmidt, M, Wilson, PA, Gutkind, JS and (LSHBCT-2005–512102). We thank Jude Samulski (University of North Chiorini, JA (2010). Epidermal growth factor receptor is a co-receptor for adeno- associated virus serotype 6. Nat Med 16: 662–664. Carolina at Chapel Hill, NC) for providing plasmid pXX6, Vyomesh 30. Humphries, JD, Byron, A and Humphries, MJ (2006). Integrin ligands at a glance. J Cell Patel (Oral & Pharyngeal Cancer Branch, National Institute of Dental Sci 119(Pt 19): 3901–3903. and Craniofacial Research, National Institutes of Health, Bethesda, 31. Jackson, T, Clark, S, Berryman, S, Burman, A, Cambier, S, Mu, D et al. (2004). Integrin alphavbeta8 functions as a receptor for foot-and-mouth disease virus: role of the beta- MD) for sharing reagents, and Hanna Janicki (University of Cologne, chain cytodomain in integrin-mediated infection. J Virol 78: 4533–4540. Germany), and Stephanie Cambier (UCSF, San Francisco, CA) for tech- 32. Mu, D, Cambier, S, Fjellbirkeland, L, Baron, JL, Munger, JS, Kawakatsu, H et al. (2002). nical assistance. The authors declare no conflict of interest. The integrin alpha(v)beta8 mediates epithelial homeostasis through MT1-MMP- dependent activation of TGF-beta1. J Cell Biol 157: 493–507. 33. Johns, HL, Berryman, S, Monaghan, P, Belsham, GJ and Jackson, T (2009). References A dominant-negative mutant of rab5 inhibits infection of cells by foot-and-mouth 1. Kubo, A, Nagao, K and Amagai, M (2012). Epidermal barrier dysfunction and disease virus: implications for virus entry. J Virol 83: 6247–6256. cutaneous sensitization in atopic diseases. J Clin Invest 122: 440–447. 34. Wang, LH, Rothberg, KG and Anderson, RG (1993). Mis-assembly of clathrin lattices 2. Simpson, CL, Patel, DM and Green, KJ (2011). Deconstructing the skin: on endosomes reveals a regulatory switch for coated pit formation. J Cell Biol 123: cytoarchitectural determinants of epidermal morphogenesis. Nat Rev Mol Cell Biol 12: 1107–1117. 565–580. 35. Van Hamme, E, Dewerchin, HL, Cornelissen, E, Verhasselt, B and Nauwynck, HJ 3. Fuchs, E (2007). Scratching the surface of skin development. Nature 445: 834–842. (2008). Clathrin- and caveolae-independent entry of feline infectious peritonitis virus 4. Gurtner, GC, Werner, S, Barrandon, Y and Longaker, MT (2008). Wound repair and in monocytes depends on dynamin. J Gen Virol 89(Pt 9): 2147–2156. regeneration. Nature 453: 314–321. 36. Eming, SA, Krieg, T and Davidson, JM (2007). Inflammation in wound repair: 5. Priya, SG, Jungvid, H and Kumar, A (2008). Skin tissue engineering for tissue repair molecular and cellular mechanisms. J Invest Dermatol 127: 514–525. and regeneration. Tissue Eng Part B Rev 14: 105–118. 37. Schechner, JS, Crane, SK, Wang, F, Szeglin, AM, Tellides, G, Lorber, MI et al. (2003). 6. Mingozzi, F and High, KA (2011). Therapeutic in vivo gene transfer for genetic disease Engraftment of a vascularized human skin equivalent. FASEB J 17: 2250–2256. using AAV: progress and challenges. Nat Rev Genet 12: 341–355. 38. De Luca, M, Pellegrini, G and Mavilio, F (2009). Gene therapy of inherited skin 7. Büning, H, Perabo, L, Coutelle, O, Quadt-Humme, S and Hallek, M (2008). Recent adhesion disorders: a critical overview. Br J Dermatol 161: 19–24. developments in adeno-associated virus vector technology. J Gene Med 10: 717–733. 39. Zhong, L, Li, B, Jayandharan, G, Mah, CS, Govindasamy, L, Agbandje-McKenna, M 8. Braun-Falco, M, Eisenried, A, Büning, H and Ring, J (2005). Recombinant adeno- et al. (2008). Tyrosine-phosphorylation of AAV2 vectors and its consequences on viral associated virus type 2-mediated gene transfer into human keratinocytes is influenced intracellular trafficking and transgene expression. Virology 381: 194–202. by both the ubiquitin/proteasome pathway and epidermal growth factor receptor 40. Zhong, L, Li, B, Mah, CS, Govindasamy, L, Agbandje-McKenna, M, Cooper, M tyrosine kinase. Arch Dermatol Res 296: 528–535. et al. (2008). Next generation of adeno-associated virus 2 vectors: point mutations in

938 www.moleculartherapy.org vol. 22 no. 5 may 2014 © The American Society of Gene & Cell Therapy Capsid-modified AAV for Cutaneous Gene Transfer

tyrosines lead to high-efficiency transduction at lower doses. Proc Natl Acad Sci USA 46. Cambier, S, Mu, DZ, O’Connell, D, Boylen, K, Travis, W, Liu, WH et al. (2000). A role 105: 7827–7832. for the integrin alphavbeta8 in the negative regulation of epithelial cell growth. 41. Grimm, D, Lee, JS, Wang, L, Desai, T, Akache, B, Storm, TA et al. (2008). In vitro and Cancer Res 60: 7084–7093. in vivo gene therapy vector evolution via multispecies interbreeding and retargeting of 47. Stachelscheid, H, Ibrahim, H, Koch, L, Schmitz, A, Tscharntke, M, adeno-associated viruses. J Virol 82: 5887–5911. Wunderlich, FT et al. (2008). Epidermal insulin/IGF-1 signalling control 42. Michelfelder, S, Kohlschütter, J, Skorupa, A, Pfennings, S, Müller, O, Kleinschmidt, JA et al. interfollicular morphogenesis and proliferative potential through Rac activation. (2009). Successful expansion but not complete restriction of tropism of adeno-associated EMBO J 27: 2091–2101. virus by in vivo biopanning of random virus display peptide libraries. PLoS ONE 4: e5122. 48. Girod, A, Ried, M, Wobus, C, Lahm, H, Leike, K, Kleinschmidt, J et al. (1999). Genetic 43. Naumer, M, Ying, Y, Michelfelder, S, Reuter, A, Trepel, M, Müller, OJ et al. (2012). capsid modifications allow efficient re-targeting of adeno-associated virus type 2. Nat Development and validation of novel AAV2 random libraries displaying peptides of Med 5: 1052–1056. diverse lengths and at diverse capsid positions. Hum Gene Ther 23: 492–507. 49. Hacker, UT, Wingenfeld, L, Kofler, DM, Schuhmann, NK, Lutz, S, Herold, T et al. 44. Ying, Y, Müller, OJ, Goehringer, C, Leuchs, B, Trepel, M, Katus, HA et al. (2010). Heart- (2005). Adeno-associated virus serotypes 1 to 5 mediated tumor cell directed targeted adeno-associated viral vectors selected by in vivo biopanning of a random gene transfer and improvement of transduction efficiency. J Gene Med 7: viral display peptide library. Gene Ther 17: 980–990. 1429–1438. 45. Uhrig, S, Coutelle, O, Wiehe, T, Perabo, L, Hallek, M and Büning, H (2012). Successful 50. Xiao, X, Li, J and Samulski, RJ (1998). Production of high-titer recombinant target cell transduction of capsid-engineered rAAV vectors requires clathrin-dependent adeno-associated virus vectors in the absence of helper adenovirus. J Virol 72: endocytosis. Gene Ther 19: 210–218. 2224–2232.

Molecular Therapy vol. 22 no. 5 may 2014 939